The overall goal of this project is to improve our knowledge concerning the optimal treatment for severely depressed inpatients. To do so, we propose to test a strategy of matching patient problems to type of psychosocial treatment. More specifically, we hypothesize that patients who receive psychosocial treatments matched their type of impairments will respond better than patients who receive treatment mismatched to their type of impairment. To test this hypothesis, we propose to subdivide a sample of depressed inpatients into four groups based on their levels of cognitive distortion and family impairment: 1) high cognitive-distortion - high family impairment; 2) high cognitive distortion - low family impairment; 3) low cognitive distortion - high family impairment; and 4) low cognitive distortion - low family impairment. All patients will receive certain common treatment elements, including hospital milieu, psychopharmacology and clinical management sessions. In addition, patients in each of the four subgroups will be randomly assigned to either a "matched" or "mismatched" treatment program. Patients in the "matched" program will receive cognitive-behavioral treatment and/or family treatment matched to the pattern in their deficits, so that patients with impairments in cognitive or family functioning will receive treatments in this area. Patients in the "mismatched" program will receive an identical amount and type of the same treatments but mismatched to their pattern of deficits, so that patients with impairments in cognitive or family functioning will not receive treatments in these areas. All treatments will begin in the hospital and will continue on an outpatient basis in active treatment for 24 weeks and on a maintenance basis for an additional 12 months.